How we work

Every entry is written by Mikhail Gavrilov, founder of BLOCKS, together with AI. He is responsible for what each entry says.

Who does what

AI finds the papers, reads the full text where it is openly available and drafts the entry. A second AI pass, which does not see the draft's reasoning, checks every number and every claim against the paper. Mikhail Gavrilov reads the result, decides what goes in and what a finding means for an athlete, and signs off the evidence level.

How an entry is made

  1. The paper is found and its DOI is matched against Crossref.
  2. The full text is read where it is open. Otherwise the entry is written from the abstract and says so.
  3. The draft gives the finding, the design, the numbers with their confidence intervals, and what the study cannot tell us. What was measured, what the authors conclude and what we infer are labelled apart; any number we calculated says so.
  4. The independent check lists every mismatch with the paper, and each one is fixed before the entry is published.
  5. The entry is linked into the guides it changes, and those guides are checked the same way.

Evidence levels

A level grades how strong the evidence in this entry is for the question it answers. It is not a verdict on the quality of the researchers' work: a careful case report is still a single case.

A+Several agreeing systematic reviews or meta-analyses, or an umbrella review
AA systematic review or meta-analysis of controlled trials, a large pre-registered trial (100+), or a major consensus statement
A-A review of mostly small or very different trials, or a well-powered trial (40+)
B+A randomised trial of 20–39 people, a crossover of 15+, a large cohort (200+), or a cross-sectional study of 100+
BA smaller trial, an uncontrolled before-and-after study, a cross-sectional or correlational study, or a small cohort
C+A single case, a narrative review, or a mechanism study with no performance outcome
CExpert opinion, animal or cell data, or a model that has not been validated

Who it was measured on

Each study says who took part, using the participant classification of McKay et al. (2022): non-athletes and recreational people, trained athletes, or elite athletes. A finding measured on untrained people may be smaller, or absent, in a trained athlete; when we apply it to athletes anyway, the entry says so.

Corrections

If you are an author, or you find something the paper does not say, use “Report an inaccuracy” on the entry. Mikhail Gavrilov reads every report. A correction is recorded in the entry's history with its date, so the change stays visible.

Older entries

The process above has been in place since October 2026. Entries written earlier are being re-checked to it, and each one is marked until it has been.

What we don't do

We don't host full texts: each entry links to the publisher, PubMed and a free copy where one exists. We don't give medical advice. No author, journal or sponsor pays to be included or to be graded.

Who pays for this

BLOCKS is a training app. The science base is free for anyone to read. The paid part is the assistant that applies it to your own training data.